Question
You provided a model called deepset/bert-base-german-cased-sentiment-Germeval17. Could you share the hyperparameters you used for training? Maybe in form of a blog entry?
Additional context
I am tryingto fit bert-base-german-cased to Germeval17 but did not reach your performance (F1 or Acc). I just changed the batch size, gradient_accumulation_steps and learning rate. What else is useful to change in an hyperparameter search?
Thanks
Philip
There is already a blog article explaining how we trained this particular model and which hyperparameters we used. You can find it here: https://towardsdatascience.com/using-transformer-based-language-models-for-sentiment-analysis-dc3c10261eec
Hope this will help you! :)
Great thanks!
Most helpful comment
There is already a blog article explaining how we trained this particular model and which hyperparameters we used. You can find it here: https://towardsdatascience.com/using-transformer-based-language-models-for-sentiment-analysis-dc3c10261eec
Hope this will help you! :)